star-bg.jpg

CognitiveScale Blog

PPP Distribution Inequities: How AI Solutions Can Eliminate Bias

By admin May 1, 2020 10:37:58 AM

0501 PPP How Bias

COVID-19 Hits Small Businesses- Hard

No doubt, given the COVID-19 pandemic, there are tough times for everyone, both personally and professionally. Enough said. Those businesses deemed non-essential and already operating on a fiscal shoestring remain in more dire straits than ever. Establishment closures hit hardest are in the small business ecosystem, where they are experiencing the greatest impact for employers and employees alike. These mom and pop and mid-sized establishments, without customers, have little recourse to help them stay afloat, let alone to assist their workforce via continued paychecks, medical coverage, and the ability to feed their kids, maintain their homes, and hopefully get back to work. 

PPP Money- Poof, It’s Gone

But let’s just pause for a moment and consider issues that are critical to some of these businesses as they attempt to stave off fiscal disaster. The nearly $350 billion Payroll Protect Program (PPP), conceived and passed by the US government to assist small businesses, was rolled out in mid-April as part of the Corona Virus Relief and Economic Security (CARES) act; the program almost immediately ran out of funds. 

PPP Distribution Inequities- A Head Scratcher

Perhaps there is no surprise as to a shortage of funds, given the scores of small businesses nation-wide applying for relief. But maybe the true surprise, and not a good one, is regarding some of the loan recipients. The PPP program is currently undergoing major controversy regarding the equitable distribution of funds. Some of America’s largest financial institutions are being challenged legally via at least four class action lawsuits contesting their choice of loan receivers, some discernibly skewed toward obtaining a more profitable return on the granted loans. This seems especially egregious with regards to access to PPP monetary resources by underserved minority businesses, ostensibly the constituents the plan was designed to assist.

Minority Businesses- They Give But Find It Hard to Get

Enterprises in communities of color have been disproportionately hard hit by the pandemic. According to the New Orleans Tribune, businesses of color (mostly small- to medium-scale) together are responsible for employing 8.7 million people and represent 30 percent of all U.S. businesses. Additionally, the combined contributions that these businesses make to the national economy is a noteworthy $1.38 trillion. It is also important to note that black-owned businesses are the second largest employer of black people in the nation–outpaced in their hiring of Black people only by the government. Still, they seem to have been passed over by PPP program funding, sparking the lawsuits and a fair amount of outrage.

Blasting Lending Bias- More Funds, Adjusted Distribution

Net-net is that lending biases are now being litigated, and almost daily a prominent organization pledges to return PPP loan money not essential to their survival. Still, even with the additional funding approved by the Feds on April 23rd it remains to be seen if the money reaches the minority businesses it is in part designed to address, now administered through smaller, more regionally focused bodies. Again, according to the New Orleans Tribune, the bill will send a total of $60 million of PPP money to small and medium-sized financial institutions to reach more black and minority-owned businesses. About $30 billion will go to community development financial institutions (CDFIs), community-based lenders, minority depository institutions (MDIs), small banks, and small credit unions. Another $30 billion will be directed to mid-sized banks and credit unions. Time, proverbially, will tell.

Does AI Boost- or Debunk- Fairness in Decision Making?

The reality of AI solutions in today’s business mix delivers a dichotomy in the overall decision making processes: Are machine-decisions more accurate than human ones, who is to decide, and what are the criteria necessary to determine ultimate “fairness” in results? With a general agreement that unconscious human influences often impact sensitive outcomes (i.e. influences of race, age, education, economic status, geography), so too similarly insufficient AI models and the datasets they employ, often based on similar baseline demographic variables, can stratify and even falsely deliver results. This might affect such criteria as the predicted ability to repay a loan, hence to qualify for federal assistance like PPP.  Unbiased decisions here can make the difference between a business’s ability to sink or swim, especially those struggling in the face of the COVAD-19 pandemic.

AI and PPP- What is the Connection?

At CognitiveScale, we develop Artificial Intelligence and Machine Learning (AI/ML) solutions that are central to eliminating discriminatory situations like those experienced by the PPP. CognitiveScale addresses the critical need for trust in AI applications. Its first-to-market data and AI/ML model vulnerability detection and risk management product, Cortex Certifai, provides six key dimensions of AI model explainability including fairness, reliability, privacy, inclusiveness, transparency, and accountability. As applied to the PPP scenario, all these dimensions could have been utilized to avoid litigation, bad publicity, and compromise to the overall objectives of the program.

CTA(1)